Gravitational search algorithm for feature selection in intrusion detection system

Abstract

This project was carried out to use the Gravitational Search Algorithm for feature selection in IDS to selectively choose significant features which represents categories of network such as DoS, Probe, U2R and R2L and to improve the accuracy and effectiveness of feature selection and to have better detection. This project aimed to study trends of feature selection in IDS and to implement BGSA for selectively choose features for IDS and to test and validate the performance and feedback of BGSA. The significance of feature selection can be viewed in two aspects. First is to filter out noise and remove redundant and irrelevant features and over load of features which causes significant loss of accuracy and time consumption in detection. In this project, it validates and evaluates the BGSA algorithm and focuses on the feature selection by implementing of BGSA. The results of BGSA program proves that the selected features which proposed by BGSA in terms of accuracy and efficiency are quite acceptable. The comparison of classification rates for all the five classes with other approaches which are using the same dataset shows that the BGSA is more accurate than others.